Source Depth Estimation Using Array Processing Techniques
Abstract
Source depth estimation is a key process in the discrimination of earthquakes and explosions for nuclear treaty monitoring. Depth estimation can be accomplished by measuring the time separation between an observation's primary arrival and associated depth phases. The lack of observable depth phases does not mean an event occurred at or near the surface. Shallow events can have closely spaced depth phases that are unperceivable to human analysts, and regional events are often complicated by the simultaneous arrival of multiple phases, which makes the observation of depth phases even more problematic. Source depth parameters for such events can be derived using moment tensor inversion or waveform modeling. These methods are, however, complicated, time consuming, and require extensive computational resources. As a result, they are not employed on a routine basis and are often limited to special event analysis. If reflected phases with sufficient signal to noise ratio reside in an observation, they will produce a spectral scalloping pattern with a period equal to the time delay between signals. This spectral phenomenon can be detected using cepstral-processing techniques, which has been used in previous studies to estimate the depth of a variety of regional and teleseismic events. However, these studies did not exploit the power of seismic arrays to determine the ray parameter of the arriving phase. We will describe a simple method for source depth estimation using array processing techniques. It requires a series of observations of a single event by a network of seismic arrays and calculates site-specific depth phase delay times and ray parameters using cepstral processing and frequency-wavenumber analysis. These measurements are used to determine the angle incidence and vertical phase velocity for each station in the observing network using simple vector decomposition. Individual depth estimates are then averaged to obtain a mean and estimate of variance. The mean is then used to recalculate the event's location at a fixed depth. An adaptive, site specific, detection threshold, derived from pre-signal noise, and a binary stacking algorithm are used to reduce the high false alarm rate inherent to cepstral processing. This process was used to estimate the depth of a shallow Bhuj aftershock whose depth phases are indistinguishable from the primary arrival. Waveform models generated from a published moment tensor solution are used to verify our result.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2006
- Bibcode:
- 2006AGUFM.S13B0224J
- Keywords:
-
- 7200 SEISMOLOGY;
- 7215 Earthquake source observations (1240);
- 7219 Seismic monitoring and test-ban treaty verification